5 ms·

Probably my favourite Lamport story is that his original Paxos paper was written in such a basic style (literally, Byzantine generals) that readers failed to re

by moomin 7y ago

Probably my favourite Lamport story is that his original Paxos paper was written in such a basic style (literally, Byzantine generals) that readers failed to realise exactly how revolutionary the content was.

SilasX 7y ago
Academic papers are almost encouraged to be hard to follow to make them sound more impressive.
- std_throwaway 7y ago
  From experience I can tell you that the solution often is very very complex in the beginning. So complex even the researcher doesn't understand it. Publication comes when the solution can be expressed in a way that at least a handful of researchers can understand it. But this is not the end. Going back and looking at it one, three, ten, twenty years later people often find better, easier ways to express it.
- lanstin 7y ago
  First third should be able to be understood by everyone even the students. Second third should only be comprehensible by professors. The last third should only be comprehended by the speaker. (Old joke about seminar talks).
- CobrastanJorji 7y ago
  Yes, but the Paxos paper is the opposite. It read like a silly homily. It's like "why's poignant guide to algorithms" except that the algorithm being described was complex, novel, and incredibly useful. It was so not full of itself that it was ignored for years. It seems to be proof that trying to sound impressive is actually important.
  - perennate 7y ago
    > algorithm being described was complex I'd argue that the algorithm itself is simple, and that the difficult part is understanding why it works (i.e., why it satisfies the safety guarantees). > It seems to be proof that trying to sound impressive is actually important. I agree with this, and would add that making a paper sound impressive rarely conflicts with making it easy to understand. If anything, pointing out the contributions and applications of an approach help readers to understand that approach. At any rate, the same paper made easier to understand will generally be more likely to be accepted.
cube2222 7y ago
Actually, the original Paxos paper is about a "forgotten Greek island, Paxos, and its Parliament". And allegedly he presented it in an Indiana Jones outfit the first time. Byzantine generals were there in one of his earlier papers.
- perennate 7y ago
  Link for anyone who wants to check it out: https://lamport.azurewebsites.net/pubs/lamport-paxos.pdf https://lamport.azurewebsites.net/pubs/lamport-paxos.pdf
- truncate 7y ago
  Yep, the original Paxos paper wasn't easy to understand, and he says so in his follow up paper "Paxos made simple"[1] whose name justifies its existence as well. [1] https://lamport.azurewebsites.net/pubs/paxos-simple.pdf https://lamport.azurewebsites.net/pubs/paxos-simple.pdf
scottlamb 7y ago
That's interesting because I had the opposite impression: important but hard to understand. I think this has been common since at least 2001, given this text on his website: > At the PODC 2001 conference, I got tired of everyone saying how difficult it was to understand the Paxos algorithm, published in [122]. https://lamport.azurewebsites.net/pubs/pubs.html#paxos-simple https://lamport.azurewebsites.net/pubs/pubs.html#paxos-simpl... I didn't start working on distributed storage systems until 2007. Perhaps some people who read the paper in 1997 had a different impression. Edit: maybe we have different takes on this story from that same url: > My attempt at inserting some humor into the subject was a dismal failure. People who attended my lecture remembered Indiana Jones, but not the algorithm. People reading the paper apparently got so distracted by the Greek parable that they didn't understand the algorithm. You're saying they didn't understand the importance; I took that to mean they didn't understand the details. ... and reading onward from the section I quoted, it sounds like your take was correct!
- dunkelheit 7y ago
  The terminology surrounding Paxos is very unfortunate in my opinion. First there is this general air of mystery which is entirely unnecessary. Next, the paper talks about "ballots" and "voting" but the actual protocol is very different from what is usually understood by voting - in the protocol everybody just votes what the leader told them! Maybe call it soviet-style voting or something.
  - furyofantares 7y ago
    I understand your complaint about the terminology and in the linked video that same confusion came up. But the terminology does make sense to me in the context of trying to reach consensus. When I’ve been a part of a group looking to reach consensus there’s an occasional ballot/vote taken to check for consensus and there isn’t a functional difference between abstaining and voting for some non-leading option. If you just say up front that there are additional logical constraints on when to abstain vs vote for the leader that make it work in a distributed system, then I think the metaphor serves to help think about it rather than confuse.
- 7y ago
ahelwer 7y ago
I think you're thinking of his paper ordering of events in a distributed system, not the Paxos paper.

by moomin 7y ago

SilasX 7y ago

Academic papers are almost encouraged to be hard to follow to make them sound more impressive.

std_throwaway 7y ago
From experience I can tell you that the solution often is very very complex in the beginning. So complex even the researcher doesn't understand it. Publication comes when the solution can be expressed in a way that at least a handful of researchers can understand it. But this is not the end. Going back and looking at it one, three, ten, twenty years later people often find better, easier ways to express it.
lanstin 7y ago
First third should be able to be understood by everyone even the students. Second third should only be comprehensible by professors. The last third should only be comprehended by the speaker. (Old joke about seminar talks).
CobrastanJorji 7y ago
Yes, but the Paxos paper is the opposite. It read like a silly homily. It's like "why's poignant guide to algorithms" except that the algorithm being described was complex, novel, and incredibly useful. It was so not full of itself that it was ignored for years. It seems to be proof that trying to sound impressive is actually important.
- perennate 7y ago
  > algorithm being described was complex I'd argue that the algorithm itself is simple, and that the difficult part is understanding why it works (i.e., why it satisfies the safety guarantees). > It seems to be proof that trying to sound impressive is actually important. I agree with this, and would add that making a paper sound impressive rarely conflicts with making it easy to understand. If anything, pointing out the contributions and applications of an approach help readers to understand that approach. At any rate, the same paper made easier to understand will generally be more likely to be accepted.

cube2222 7y ago

Actually, the original Paxos paper is about a "forgotten Greek island, Paxos, and its Parliament". And allegedly he presented it in an Indiana Jones outfit the first time. Byzantine generals were there in one of his earlier papers.

perennate 7y ago
Link for anyone who wants to check it out: https://lamport.azurewebsites.net/pubs/lamport-paxos.pdf https://lamport.azurewebsites.net/pubs/lamport-paxos.pdf
truncate 7y ago
Yep, the original Paxos paper wasn't easy to understand, and he says so in his follow up paper "Paxos made simple"[1] whose name justifies its existence as well. [1] https://lamport.azurewebsites.net/pubs/paxos-simple.pdf https://lamport.azurewebsites.net/pubs/paxos-simple.pdf

scottlamb 7y ago

That's interesting because I had the opposite impression: important but hard to understand. I think this has been common since at least 2001, given this text on his website: > At the PODC 2001 conference, I got tired of everyone saying how difficult it was to understand the Paxos algorithm, published in [122]. https://lamport.azurewebsites.net/pubs/pubs.html#paxos-simple https://lamport.azurewebsites.net/pubs/pubs.html#paxos-simpl... I didn't start working on distributed storage systems until 2007. Perhaps some people who read the paper in 1997 had a different impression. Edit: maybe we have different takes on this story from that same url: > My attempt at inserting some humor into the subject was a dismal failure. People who attended my lecture remembered Indiana Jones, but not the algorithm. People reading the paper apparently got so distracted by the Greek parable that they didn't understand the algorithm. You're saying they didn't understand the importance; I took that to mean they didn't understand the details. ... and reading onward from the section I quoted, it sounds like your take was correct!

dunkelheit 7y ago
The terminology surrounding Paxos is very unfortunate in my opinion. First there is this general air of mystery which is entirely unnecessary. Next, the paper talks about "ballots" and "voting" but the actual protocol is very different from what is usually understood by voting - in the protocol everybody just votes what the leader told them! Maybe call it soviet-style voting or something.
- furyofantares 7y ago
  I understand your complaint about the terminology and in the linked video that same confusion came up. But the terminology does make sense to me in the context of trying to reach consensus. When I’ve been a part of a group looking to reach consensus there’s an occasional ballot/vote taken to check for consensus and there isn’t a functional difference between abstaining and voting for some non-leading option. If you just say up front that there are additional logical constraints on when to abstain vs vote for the leader that make it work in a distributed system, then I think the metaphor serves to help think about it rather than confuse.
7y ago

ahelwer 7y ago

I think you're thinking of his paper ordering of events in a distributed system, not the Paxos paper.